From September 1, 2025, all AI-generated content published in China will have to be clearly labeled.
Four administrative bodies in China - Cyberspace Administration of China, Ministry of Industry and Information Technology, Ministry of Public Security, State Administration of Radio and Television - have announced a regulation on mandatory labeling of AI-generated content. The regulation has 14 articles and specifies who and what is subject to its requirements
👨⚖️ The regulation applies to texts, images, audio, video and virtual scenography
👨⚖️ AI content identification is to be both direct and indirect
👨⚖️ For text - marking with text or symbol at the beginning, middle and end of the text or adding visual markings in the interface of the interactive scene or around the text
👨⚖️ For audio - voice warning - as above
👨⚖️ For image - visible marking indicating AI
👨⚖️ For video - visible marking at the beginning, middle and end of the video
👨⚖️ In all cases, the content metadata must contain information that it was generated by AI
👨⚖️ If the above content has the option to be copied, downloaded or exported, the files must contain appropriate information indicating that it was generated by AI
👨⚖️ Internet service providers are responsible for ensuring that the content they publish is correctly marked as being generated by AI. They do this based on: a) metadata entries; b) a declaration by the content provider if there is no clear description in the metadata; c) a statement by the provider that it is AI-generated content even if there is no mention of this in the metadata or the provider's declaration. 👨⚖️ Information on the methods of marking AI-generated content must be clearly described and available to every user of the service providers
👨⚖️ It is prohibited to remove, modify or hide AI-generated content markings
Content of the Notice on the Issuance of Measures for Identifying Synthetic Content Generated by Artificial Intelligence (Guoxinbantongzi [2025] No. 2) (in Chinese)
https://www.gov.cn/zhengce/zhengceku/202503/content_7014286.htm
I will add that I have the impression that these regulations are more stringent than the EU regulations.
The regulations primarily concern AI tool developers, owners of internet platforms, and creators of advertisements and content. The former have many reservations about the introduced regulations, because they will be an obstacle to competing with Western companies that do not experience such restrictions.
Yes, Chinese regulations on AI-generated content are generally stricter in terms of state control and censorship, while EU regulations focus more on transparency, accountability, and fundamental rights. China’s rules—such as the 2023 “Interim Measures for Generative AI Services”—require prior content review, adherence to socialist values, and real-name verification. In contrast, the EU’s AI Act emphasizes risk-based regulation, requiring disclosure for synthetic content and safeguards for human rights, but allows broader innovation space. Thus, China prioritizes political control, while the EU balances regulation with civil liberties and innovation.
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India